Diffutron: A Masked Diffusion Language Model for Turkish Language

TL;DR

Diffutron is a resource-efficient masked diffusion language model tailored for Turkish, demonstrating competitive performance in non-autoregressive text generation through a multi-stage training process.

Contribution

Introduces Diffutron, a novel masked diffusion language model for Turkish, utilizing a multi-stage training pipeline with continual pre-training and instruction tuning.

Findings

Achieves competitive results with smaller model size

Effective multi-stage training improves Turkish language modeling

Validates masked diffusion approach for morphologically rich languages

Abstract

Masked Diffusion Language Models (MDLMs) have emerged as a compelling non-autoregressive alternative to standard large language models; however, their application to morphologically rich languages remains limited. In this paper, we introduce $\textit{Diffutron}$, a masked diffusion language model specifically designed for Turkish. Our approach leverages a resource-efficient training pipeline, starting with LoRA-based continual pre-training of a multilingual encoder on a large-scale corpus. To enable generative capabilities, we employ a progressive instruction-tuning strategy, sequentially adapting the model on general and task-specific instruction sets. Experimental results across comprehensive benchmarks demonstrate that, despite its compact size, our model achieves competitive performance compared to existing multi-billion-parameter baselines. These findings validate the effectiveness of masked diffusion modeling combined with multi-stage tuning for non-autoregressive text generation in Turkish.